Elisa Zhang
Dougherty Valley High School
San Ramon, California
Patching Multi-Location Bugs: A Multi-Agent Large Language Model Framework for Automated Code Repair
Elisa developed an AI-based system to automatically patch bugs that span many locations in software programs.
View Poster
Elisa Zhang, 17, of San Ramon, developed an AI system to automatically patch software bugs across multiple locations for her Regeneron Science Talent Search computer science project. Multi-location bugs can span thousands of lines of code. They plague real-world programs and are hard to address. Unfixed bugs can cause problems ranging from system crashes to security breaches, costing billions of dollars a year.
In her project, Elisa programmed unique large language model (LLM) agents with individual prompts and contexts. The LLM agents worked together to code, verify and select the best possible fix. Elisa tested her system on a dataset of 190 Java bugs, and it outperformed existing repair methods. She believes her system can simplify bug fixes.
Elisa, the child of Qing Zhang and Qian Zheng, attends Dougherty Valley High School, where she is an officer of the math club. As external vice president of the Promoting Leadership in Aspiring Youth (PLAY) nonprofit, she leads the PLAY Times newsletter, helps manage the website and is co-president of its math club.
Beyond the Project
Elisa presented research at computer science conferences in 2023 and 2024. At both conferences, she was the only high school student presenting work alongside scientists.
FUN FACTS: Elisa loves to dance. She is a competitive dancer who performs traditional Chinese, lyrical and contemporary dances. She holds her dance studio’s record for the most fouetté turns — 60!
